What makes a great oil spill forecast?

The oil and shipping industries are committed to carrying out safe and reliable operations with the goal of preventing oil spills. Significant resources are dedicated to achieving this goal and the number of oil spills continues to decrease thanks to such efforts (ITOPF, 2015). Despite best efforts, oil spills cannot be avoided completely. If a spill occurs, the impact on the environment can be minimised through a rapidly executed and well-managed response strategy.

To create an effective response strategy, Incident Managers need to know what the fate of the spilled oil is likely to be. They will want the answers to questions that can be broken down into two key themes: where is oil likely to travel and how could oil physically transform?

Information obtained from oil spill models has been used to answer these and other questions for many years (Johansen, 1987). For example, by using forecast metocean data such as ocean currents and wind, modellers can predict where spilled oil will travel in the next few days. Combining this with Geographic Information System (GIS) data allows the modeller to determine which jurisdictions and sensitive resources are likely to be affected and whether or not shoreline impact could occur. Characteristics of the spilled oil, such as API, viscosity, wax and asphaltene content and pour point, are used to match the spilled oil to an analogue in a model's database. These characteristics are the main factors in determining the weathering oil will undergo, for example, will oil take on water and emulsify?

As modelling can help answer many urgent questions Incident Managers and other stakeholders have during a spill, it is an invaluable component of the surveillance, modelling and visualisation toolbox. In-field, aerial and satellite surveillance are important tools for determining the current status of an oil spill, and they are fed back into oil spill models to improve the forecasts as the spill develops.

There are many papers already published which discuss the science behind various oil spill models and their performance, such as Hong et al, 1997 and Daling et al, 1997. This paper looks at the communication of an oil spill forecast and how modellers can ensure products produced are effective and easy to use during an incident.

Oil spills are dynamic and therefore forecasts are well suited to delivery methods that can show change over time such as videos or as a component of a Common Operating Picture (COP).¹ These dynamic deliverables should be supplemented with a report so the end user can understand the prediction and be used as an audit trail. By following five simple steps we can achieve our goal of an oil spill forecast that is effective and easy to use during a spill.

Large reports (well in excess of 50 pages) are often created using the data derived from a model run of just a few days. This is especially true if the modeller includes details of how the model works; a complete set of model outputs and other information that may not be deemed of high importance during a response. Put yourself in the shoes of an Incident Manager...

Incident Managers, and the vast majority of others responding to an oil spill, do not want to receive long reports full of detail that is not relevant to them during a spill. For example, the Incident Manager is unlikely to be interested in the details of the metocean datasets, such as temporal and spatial resolutions, used to run an oil spill model; a simple reference to the datasets will suffice. A useful rule of thumb is to keep reports to one or two pages of information that can be read within ten minutes. Time is a precious commodity during a spill!

In addition to keeping the technical sections of the report clear and concise, good version control is also essential. Each report needs to be clearly dated so the end user knows they are working from latest forecast.

In table 1, we looked at some of the questions an Incident Manager may have during a response, such as where will oil go and what will happen to its physical properties?

As with most geographic data, the best way to describe where oil could go is by using a map. Maps are effective at conveying a large amount of data at once. As an oil spill is dynamic there are a large number of maps the modeller could create. Good quality oil spill forecasts are restricted to around five days into the future due to difficulties predicting medium and long range weather. Assuming a map is produced for each hour of the spill, there would be 120 maps for a five day forecast which clearly does not lead to a concise report. One method is to present a single map which shows the key data for the full length (say 5 days) of the forecast. A simple way to do this is by showing a map with the footprint of the spill at the end of the forecast which includes the area oil has travelled through. This can be supplemented with a video or another dynamic deliverable, such as a shapefile, to show how the spill develops over time.

An Incident Manager or indeed any key stakeholder will want to know if oil is likely to reach sensitive resources or cross into other countries waters. By adding GIS data, such as sensitive areas of maritime boundaries, to modelling results maps this information can be obtained by a quick glance at the report.

Properties of oil changes quickly once it is spilled into the marine environment. The extent of this will depend heavily on the type of oil spilled but in most cases oils will become heavier, more viscous and may take on water to emulsify over time. Incident Managers will be interested in this as they will want to know what response techniques are most suitable. For example if the viscosity of the oil exceeds 10,000 centistokes (the consistency of honey) chemical dispersants may not be effective. The oil's viscosity will also be significant when deciding what type of pumps and skimmers to use. Incident Managers will want to know how the oil properties will change over time so line graphs are effective ways of conveying this information.

Keeping the question in mind also means thinking about why the question was asked. An Incident Manager may ask “How much waste do you think will be available for collection from the sea surface after two days?” Oil spill modellers can help estimate this. A common mistake is to interpret this as the volume of oil that will be on the sea surface at the end of day 2. As many oils emulsify by taking on water, the volume of waste can be increased by up to four fold. This is a confusing concept that can be easily mis-communicated during a spill. Therefore it is good practice to report the volume of oil and the volume of emulsion, with an explanation if necessary. The volume of oil is often reported using a mass balance plot which shows the volume of oil: on the surface; in the water column; on the shore and in the atmosphere. Emulsion cannot be plotted on a mass balance plot as oil dispersed in the water column and components that have evaporated to the atmosphere do not emulsify. An emulsion plot, for surface or shoreline waste, will typically be presented alongside a mass balance plot.

The detail of the explanation that accompanies the emulsion plots will depend on the end user. Incident Managers and responders are well versed in the emulsification of spilled oil so will only need a limited explanation or reminder. A regulator, who is likely to be involved in a range of issues in addition to oil spills, may need a more detailed explanation. A member of the public, with limited or no oil spill experience will require a more detailed explanation.

So far we have focused on Incident Managers and Responders but there are other stakeholders who may require access to an oil spill modelling forecast. Government officials, the responsible party, local businesses and members of the public will be interested in seeing these products.

Many of the larger operators have their own modelling teams whereas others subcontract to response organisations. The responsible party should have a good knowledge of oil spills and will be working closely with response organisations. Their requirements are likely to be similar to that of an Incident Manager – in fact many operators will have their own Incident managers and responders.

Government official's knowledge of oil spills will vary greatly depending on the training they have received and the experience they have of previous spills and exercises. Some government officials are unlikely to specialise in oil spills, in the way responders and responsible parties do, the reports they are given will need more explanation. Government officials are most likely to be interested in where the oil is going and what resources will be affected. They will need to know if oil is likely to cross into the surface waters of other countries.

An oil spill can spark interest across the globe as well as in the local community that is likely to be impacted. As the majority of the public do not have any oil spill training, communication of modelling results needs to tailored accordingly. For example, we do not need to provide information on emulsions, viscosity, groups of oil, resolution of metocean datasets and other specialist ideas to members of the public. It is much better to give a simple summary of the development of a spill concentrating on where oil is likely to go and how the cleanup is progressing.

Ideally a different modelling report will be issued to the various stakeholders. The modeller will only need to run one set of models but the interpretation and communication of the report should specific to the end user.

When writing a short report (1–2 pages) a summary of the key findings is sometimes overlooked. As discussed earlier in the paper an oil spill forecast should be simple and concise because of the time pressures associated with an emergency incident. This makes summarising the key information vitally important as during busy parts of the response the end user may only quickly glance at the forecast.

Summarising the key information can be thought of as its own mini report and it should consider the steps we have already considered:

• Keep it short and concise

• Focus on the questions

• Think about the end user

In the summary the modeller should just focus on the key questions that the end user wants answered. The summary will usually be descriptive – a few bullet points which conveys the key information from the maps, plots and statistics is ideal.

Oil Spill models attempt to predict an extremely complex natural event and there is inherent uncertainty within them due to factors such as imperfect algorithms, lack of observations and poorly trained users. The summary box is a useful place to flag any key factors of uncertainty.

Following the previous four steps will help a modeller produce an effective oil spill forecast that is easy to use. This final step will not help the end user actually make predictions about the spill but it will help them or others repeat the model setup if necessary. Models will need to be reproduced and tweaked for a number of reasons. For example, surveillance may show oil moving in a different direction; the release rate may have been underestimated or more recent hydrodynamic forecasts may be available.

Responsible parties and government bodies often have their own modelling capability. They may wish to compare models and check what has been produced. To do this reference to the input data will be needed. Key inputs include metocean data (currents and winds), oil type, release location, model duration, release time and release rate. The quality of these inputs is essential to the accuracy of the predictions. The modeller should include information on the model inputs in the report and give enough detail for another modeller to be able to reproduce the results but this should be done without overcomplicating the report and detracting from its focus on the end user.

An oil spill forecast is very useful whilst developing a response strategy. Forecasts are used to predict where a spill will go and can assist in determining what response techniques will be best suited to minimising its impact on the environment. Poorly communicated forecasts can be long, difficult to read and even mis-leading. It is therefore vitally important that oil spill modellers work hard to communicate their forecasts in an effective manner that makes them easy to understand in a high pressure environment. Keeping reports short and concise, focusing on the questions, thinking about the end user, summarising the key information and ensuring the forecast is repeatable will result in a great oil spill forecast that really aides the response.